Session: 37-01 Machine Learning and Optimization

Paper Number: 81091

81091 - Multi-Objective Development of Machine-Learnt Closures for Fully Integrated Transition and Wake Mixing Predictions in Low Pressure Turbines 

Reynolds averaged Navier Stokes (RANS) and unsteady RANS flow simulations in low pressure turbines (LPT) for aeroengine applications still suffer from some severe limitations concerning transition and wake mixing predictions. Due to the low Reynolds number a large part of the blade boundary-layer remains laminar and transition may occur due to flow separation. The boundary-layer details at the blade trailing edge can change substantially depending on the transition region topology and this strongly influences the wake mixing occurring downstream, so that such flow features appear as strongly coupled in LPT flow environments. In this work a recently developed computational fluid dynamics driven machine-learning framework coupled with multi-expression, multi-objective optimization is exploited to generate transition and turbulence closures aimed at improving both transition and wake-mixing predictions in LPTs. The T106A blade cascade is adopted as a training case, where the considered Reynolds numbers range from 60,000 to 100,000.  

The baseline transition model is based on a laminar kinetic energy transport approach, and the machine-learning approach is used to reformulate the source terms as functions of suitably defined non-dimensional ratios. Additionally, machine-learning based explicit algebraic Reynolds stress models are used to improve wake-mixing predictions. A sensing-function based strategy is devised to allow an automated zonal application of the developed models. 

Presenting Author: Harshal Akolekar Univeristy of Melbourne

Presenting Author Biography: Harshal Akolekar is a postdoctoral fellow at the University of Melbourne, Australia. He is currently working on developing machine-learnt turbulence / transition closures for a range of flows relevant to gas turbine aerodynamic and cooling applications. He worked as a research scientist in the Maritime Division, at the Defence Science and Technology Group (DSTG), Australia, in the area of submarine hydrodynamics from 2019-2021. He received his PhD from the University of Melbourne in 2019 in the area of data-driven turbulence model development for low pressure turbines in collaboration with General Electric (GE), Aviation.

Authors:

Harshal Akolekar Univeristy of Melbourne
Fabian Waschkowski University of Melbourne
Roberto Pacciani Univeristy of Florence
Yaomin Zhao Peking University
Richard Sandberg University of Melbourne